Gadgetory

aside from their Bauer there's one major problem with extreme overclocking and liquid nitrogen benchmarking and that is it's all done with synthetic benchmarks like x bi which are great for competitive scoring and seeing how the cards plays but what we're doing today is gaming benchmarks so we were writing the kingpin 2080 ti through our actual benchmark suite for GPUs including real games like Sniper Elite for F 1 2018 far cry 5 shadow of the Tomb Raider and we're doing it while it's under liquid nitrogen so we're going to be mounting an L n to pot to the card going through the benchmarks trying to maintain the pot temperature and see how much does it scale in reality when you're at 25 65 megahertz before that this video is brought to you by thermal grizzly and their high-end thermal compounds Thermal Grizzly makes cryo not paste for high thermal performance and conductivity without being electrically conductive so you don't have to worry about shorting components cryo knot is particularly good for replacing stock GPU pastes as cryo knot is a non curing compound learn more at the link in the description below the most common complaint we've gotten in the liquid nitrogen overclocking streams is that apparently people think it's not practical or something so we're gonna show you how easy it really is all you need is a three to four hundred dollar liquid nitrogen pot ideally you also get a $1,900 video card that is pre-prepared for liquid nitrogen although you could do some hard mod yourself you'll need a giant tub of Vaseline a brush some good paste for - 130 or below degree Celsius temperature so we're using kpx for that need a heat gun and that some liquid nitrogen so this is actually something everyone can do at home please actually look into what you're doing if you if you try to follow that advice so yeah it's it's obviously impractical but what we're gonna do is set it up and do some gaming with L on to anyway and all we need to do for this is clean off the card get vaseline all over the memory on the vrm there's already some applied will heat it up and doing all of this will prep us for running the card at around minus 130 degrees Celsius this one likes to run we're going to be using settings of about 25 65 megahertz or thereabouts for the core and we'll have a setting of about 1500 megahertz offset for the memory then voltage will be in the range of about one point three to five to one point four four V core and we'll be running about one point five for the memory voltage and NV VDD OCP will be turned off and that should be about all we need so then finally to answer the comment of liquid nitrogen overclocking is is pointless because you can't play games we will now raise you with actually yes you can and all we have to do is have two people one to manage the pot temperature and one to play the game and that will be everything required for the ln2 gaming benchmarks and then we'll get into the numbers so what we'll be doing is we're going to set Patrick up to play a game I will be managing the liquid nitrogen temperature and pouring more ln2 between the two of us we should be able to actually game while on lnto so we'll have some numbers in the actual charts in benchmarks the exact same way as every other GP that we have on the charts and they'll get you get you all the information so we can see the actual scaling while we're doing gaming benchmarks on something that's at say 500 megahertz higher than the actual stock cards are really meant to run that'll show us once and for all what kind of scaling there is if any in real gaming workloads and not just in synthetic workloads Sniper Elite 4 starts us out this game uses DirectX 12 and the best implementation we've seen to date for a game and has known scaling to avoid CPU bottlenecks on our platform we also use asynchronous compute which is well leveraged on modern architectures the kingpin 20 atti at twenty five sixty five megahertz core and plus fifteen hundred memory putting it about five hundred megahertz more than most overclocked twenty eighty TI's ends up at 148 FPS average with extremely well times low performance in fact the Ellen 2 K P card outperforms the water old KP overclocks 131 FPS average by an impressive 13.5 percent moreover it significantly outperforms the borderline instability experienced in our water-cooled overclock compared to the stock kingpin card at 120 FPS average the improvement with ln2 is 24% with the frequency increase about 26% over the 20 40 megahertz baseline this is a fairly linear scaling and shows impressive gains although a large part of that is attributable to the well-built nature of Sniper Elite 4 interestingly the NV link 20 atti cards end up at about 210 FPS average posting nearly perfect scaling but the KP 20 atti splits the difference when under l on - between the NV link and the stock single card results not bad here's where it gets interesting this frame time plot shows the kingpin 20 atti stock card performance plotted in blue with the ln 228 ET ikp card reaching frame times closer to 6 milliseconds that's impressive remember lower is better here and the consistency is even more impressive we don't encounter any excursions from the mean equal to or greater than 8 milliseconds so the user gets functionally a perfect fluid experience at around 150 FPS average with good frame times for reference sixteen point six six seven milliseconds would be 60 FPS average which we are well below but the KP 20 80 TI with water cooling and a borderline unstable overclock encountered more excursions that manifest as small stutters in gameplay clearly gaming on Allen 2 is both very practical and objectively superior and everyone should immediately by Allen - Potts doers and switch to Allen - gaming rigs with co-drivers for pouring f1 2018 is next giving us a DirectX 11 title to look at for this one the only setup that beats our ln 2 KP card is the dual card Titan RT x configuration with NV link running a 175 FPS average with frankly abysmal frame time consistency the KP 20 atti at 25 65 megahertz and + 1500 megahertz memory runs at 138 FPS average meaning the dual Titan RT axes lead only by 26% despite having a financial requirement nearing 3 times higher versus the water-cooled KP overclocked at 122 FPS average our increase is an impressed 14% was a 0.1% low frame time performance doubled compared to the stock 20 ATT I can been at 113 FPS average the increase is 23% and it's about 31% faster than the 28 ECI XC Ultra is 106 FPS average shadow the tomb Raider's next swinging us back to DirectX 12 for this one the ln2 cooled kingpin 20 atti 2565 megahertz core and again a boosted memory gets us to 92 FPS average for an improvement over the previous overclocks 82 FPS average by 13% this is a consistent uplift it seems although the lows are much improved over the previous overclock clearly we have once again proven that liquid-nitrogen gaming is in fact the only way to achieve high refresh gaming and we hope that all esports arenas immediately change the cooling systems to ensure no dropped frames to the credit of the NV linked cards here the scallion is sort of insane with the Titan RT X at 138 FPS average and the xx atti at about 147 FPS average when running on its higher clocks and better coolers then we had to tighten our tea ax running even with Ellen to the KP 2080 TI has a long distance to go to be a chart-topper but it's certainly chart tops for single card configurations GTA 5 is still one of the most popular games on Steam and so we thought the GTA v community would be happy to know the completely realistic and attainable frame rates with liquid nitrogen gaming for this game tested at 4k with to tap msaa and very high to ultra settings we measured the Ellen - Kane Ben 20 atti at 97 FPS average with it well timed at low frame time performance the uplift over the water-cooled xx atti kingpin overclocked is about 11.5 percent moving from 87 to 97 FPS average compared to the stock 20 atti kingpin card and it's 20 40 megahertz average clock which is already on the level of an average over clocks 20 atti we end up improves by about 21 percent versus the 80 FPS average baseline far cry 5 is last for this one tested again at 4k the 20 ATT I can pin under liquid nitrogen ends up at 103 FPS average approaching the NV linked 20 ATT is although NV length doesn't scale as cleanly here as in a title like Sniper Elite for the 103 FPS average puts us improved over the h2o over clocks 92 FPS average by 12% sticking close to the improvement we saw previously versus the stock kingpin card we've boosted by 23% with Ellen - Ellen - Gaming was a lot of fun it actually the results were surprisingly good - thirteen percent on average improvement on - versus h2o overclocks and then we saw an increase of about 23 to 25 percent over the baseline stock numbers for the already high stock locked kingpin card so overall successful and the actually important thing here was just to see how does the scaling apply to the real world and if frequencies keep going up in a hypothetical scenario we can now look at how Torian would scale and potentially future architectures but realistically it's just kind of fun and this is not something you should be doing at home but if you wanted to get into liquid nitrogen overclocking for competitive reasons then there's plenty of resources on how to do it you only a doer you'll need Ellen - you absolutely need to read up on safety it is fairly safe to do but there are definitely things that I mean you could die if you do it wrong like it is a real actual hazard you could die so just make sure you read about what you're doing so you don't do anything stupid but yeah a lot of fun to see the results Sniper Elite had some very good scaling and that is a common trend for that particular title when we were playing when Patrick was messing around in GTA 5 we were we actually had to increase the graphics settings because we were running into the issue where we hit the GTA 5 frame rate cap of 180 7.5 FPS at which point it will stutter back really hard on specific CPUs it's way to increase the graphics settings for that reason and then frost build-up is was not too difficult to manage you just have to keep it cold and don't let it warm up which means finding a stable overclock quickly and leaving it there and for the overclock settings we ended up at about 1.3 1.3 to 5 for NV VDD and then 1.5 fpv GD and we left packs alone actually we did leave that one stock elites at one point of 87 for the rest OCP was off and any other setting oh yeah we tried overclocking Patrick's Mouse while he was using it - right i did´t i don't know that he didn't exactly ask for it but we did overclock the mouse with on to so yeah that's a that's gonna be it let us know what you think we are considering doing a live stream of this once we get back from Computex so if that's something you'd like to see let us know subscribe you go to store like Aaron's Nexus dotnet to help us out directly by buying a shirt like this one this is the GPU artifacting shirt which is particularly relevant if you're doing liquid nitrogen overclocking and you go to patreon.com/scishow Nexus tell us out there as well but the most important thing is check back for our computation coverage we will be in Taiwan for a couple of weeks doing coverage in of the show and of areas around the show thanks for watching I'll see you all next time